Clinically-aligned ischemic stroke segmentation and ASPECTS scoring on NCCT imaging using a slice-gated loss on foundation representations

TL;DR

This paper introduces a novel deep learning framework that leverages foundation model representations and structured clinical priors to improve ischemic stroke segmentation and ASPECTS scoring on NCCT images, achieving higher accuracy without added inference complexity.

Contribution

It proposes a clinically aligned segmentation framework combining a frozen foundation model with a Territory-Aware Gated Loss to enforce anatomical consistency during training.

Findings

Achieved a Dice score of 0.6385 on AISD, surpassing previous CNN and foundation-model baselines.

Improved mean Dice from 0.698 to 0.767 on proprietary ASPECTS dataset with TAGL.

Demonstrated that structured clinical priors enhance NCCT stroke segmentation accuracy.

Abstract

Rapid infarct assessment on non-contrast CT (NCCT) is essential for acute ischemic stroke management. Most deep learning methods perform pixel-wise segmentation without modeling the structured anatomical reasoning underlying ASPECTS scoring, where basal ganglia (BG) and supraganglionic (SG) levels are clinically interpreted in a coupled manner. We propose a clinically aligned framework that combines a frozen DINOv3 backbone with a lightweight decoder and introduce a Territory-Aware Gated Loss (TAGL) to enforce BG-SG consistency during training. This anatomically informed supervision adds no inference-time complexity. Our method achieves a Dice score of 0.6385 on AISD, outperforming prior CNN and foundation-model baselines. On a proprietary ASPECTS dataset, TAGL improves mean Dice from 0.698 to 0.767. These results demonstrate that integrating foundation representations with structured clinical priors improves NCCT stroke segmentation and ASPECTS delineation.